Connect sandbox to prod-clone didn't retain the edxapp version even though
it asked the user for it in the jenkins job.

Ansible will use the correct ssh params so that we don't have to worry about host key checking
which will become a problem as we build sandboxes at the same domain more than once.

Also check the return code to make sure it succeeded.

Don't override the versions is server-vars.yml.

```
13:08:11 + ret=0
13:08:11 + echo 0
13:08:11 0
13:08:11 + '[' 0 -ne 0 ']'
13:08:11 Build step 'Virtualenv Builder' marked build as failure
13:08:11 [ssh-agent] Stopped.
13:08:11 Finished: FAILURE
```

The example above is a snippet from jenkins which the integration test runs into,
meaning it is often suceeding but reporting it as failure.  There is still an issue
with the cms worker which is related to 'stopwaitsecs' which is really high for all
workers due to grading.  That work will need to be refactored so that cms workers
don't wait so long before killing the worker.

The problem is the same as what is described here but more investigation will be
useful.

http://stackoverflow.com/questions/9971627/why-is-celery-not-shutting-down-cleanly

This setting gets propogated to the launch configuration and is currently
causing unexpected behaviour.

Sometimes we get an instance id back from the creation call but then
get an exception saying the ID was not found when we do a get_all_instances.